More details were explained by Lee et al. [7].The proposed vision-based system can support two cameras at the subsystem level with 30 frames per second (fps), so it can be appropriate to track the motion of civil structures with the maximum frequency of less than 15Hz (Nyquist frequency). In phase 3 reality, civil structures usually have a natural frequency of lower than 4-5Hz [8], thus the proposed system can be applicable to measuring the dynamic displacement of large-scale structures. However, for the purpose of real-time processing, the processing time per frame should be less than 33.3 milliseconds (ms). To verify the performance of the proposed system, processing time per frame at a subsystem PC was checked as shown in Figure 9. A total of 2300 frames were captured by camcorders and processed at the subsystem.

The averaged and maximum processing time per frame are 17ms and 21.5ms, respectively, which is fast enough for real-time processing.Figure 9Processing time per frame.3. Experimental VerificationsSeveral laboratory tests were conducted to verify the proposed multipoint vision-based system and the time synchronization algorithm. To ensure good data transaction between the slave subsystems and the master PC, all the PCs and their components should be stable without viruses and spywares before conducting the verification tests. 3.1. Shaking Table TestIn order to evaluate the performance and the stability of the proposed synchronized multipoint vision-based system, several shaking table tests were performed. Figure 10 shows the shaking table and the target used in the experiments.

The results of this system were compared with those measured using LVDT. Figure 10Shaking table and target size.The testing system consisted of two laptops, one Lenovo-R61 (Intel Core Duo 2.4MHz, 2GB of RAM), and one Acer-Asprire 5580 (Intel Core 2-Duo 1.66MHz, 3GB of RAM), Lenovo was used as the slave PC. Two JVC GZ-MS120 camcorders with an optical zooming capability of 40 times, a resolution of 640 �� 480 pixels, and a frame rate of 30 fps were used. In addition, GSK-3 two myVision USB frame grabbers, two telescopic lenses, and one wireless LAN router complying with the 802.11g wireless standard were implemented to transfer data between the master PC and slave PCs. The target size is 15mm in vertical and horizontal directions. The camera was placed at 16 meters apart from the target. The number of pixels between the uppermost white spot and the lowermost white spot was 284, thus the physical resolution was 0.053mm/pixel. The schematic of the testing configuration is given in Figure 11.Figure 11Experimental location setup.